Belajar QA Automation Engineer - Agentic AI Testing

#Pendahuluan

Setelah di episode 21 kita mempelajari AI-assisted testing, pada episode ini kita melangkah lebih jauh: agentic AI testing — AI agents yang secara autonomous mengeksplorasi dan menguji aplikasi.

#Apa itu Agentic Testing?

text

Agentic Testing: AI agents yang:
├──自主 explore aplikasi
├── Identify potential issues
├── Generate test cases secara otomatis
├── Execute tests tanpa human intervention
└── Report findings dengan context

#Perbedaan dengan Traditional Automation

Aspek	Traditional	Agentic
Test cases	Ditulis manual	AI-generated
Exploration	Scripted paths	Autonomous
Adaptation	Fixed locators	Self-healing
Coverage	Known paths	Discovery-based
Maintenance	Manual	AI-maintained

#AI Agent Testing

#Browser-Use Agent

typescript

// Concept: AI agent yang mengontrol browser
const agent = new BrowserAgent({
  goal: "Find all forms and test input validation",
  maxSteps: 50,
  screenshots: true,
});
 
const results = await agent.run();
console.log(results.issues);

#Natural Language Test Cases

text

Natural Language Test:
"Given I am on the login page
 When I enter invalid email format
 And I click login button
 Then I should see email validation error"
 
→ AI convert ke Playwright code otomatis

#Visual Verification

typescript

// AI visual verification
const aiVerifier = new AIVisualVerifier();
 
// Compare screenshots dengan AI understanding
const diff = await aiVerifier.compare(
  'baseline.png',
  'current.png',
  { semantic: true }  // Understand context, bukan pixel-perfect
);
 
// AI mengerti: timestamp berubah = acceptable
// AI mengerti: button hilang = bug

#Autonomous Exploration

#Crawl & Test

typescript

// AI crawler yang mengeksplorasi aplikasi
const crawler = new AICrawler({
  startUrl: 'http://localhost:3000',
  depth: 3,
  actions: ['click', 'fill', 'submit'],
  ignorePatterns: ['/logout', '/delete'],
});
 
const siteMap = await crawler.crawl();
// Menghasilkan site map + test cases untuk setiap page

#2026 Landscape

text

Agentic Testing 2026:
├── Self-healing tests standar
├── AI agents untuk regression testing
├── Natural language test authoring
├── Visual AI verification
├── Autonomous bug discovery
├── Predictive test selection
└── AI-human collaboration workflows

Note

Agentic testing masih berkembang. Gunakan sebagai complement, bukan replacement untuk human-written tests. Human judgment tetap diperlukan untuk critical flows.

#Praktik: AI Agent Workflow

text

Agentic Testing Workflow:
1. Define goals (apa yang ingin diuji)
2. Deploy AI agents untuk explore
3. AI generate test cases
4. Review AI-generated tests
5. Integrate ke existing suite
6. Monitor dan refine

#Penutup

Agentic testing: AI agents yang autonomous mengeksplorasi dan menguji.
Self-healing: locator otomatis repair saat berubah.
Visual AI: semantic comparison, bukan pixel-perfect.
2026 trend: self-healing, autonomous, natural language tests.

Di episode 23 selanjutnya kita akan membahas testing AI/LLM applications — testing AI apps, prompt validation, dan evaluation metrics. Sampai jumpa di episode 23!